Source of Economic Growth in Ethiopia: An Application of Vector Error Correction Model (VECM)

Transcription

1 Source of Economic Growth in Ethiopia: An Application of Vector Error Correction Model (VECM) Khalid Y. Ahmed and Yokoyama Kenji Ritsumeikan Asia Pacific University (APU), Japan ABSTRACT The primary objective of this research is to examine the recent impressive economic growth of Ethiopia and to evaluate the major determinates of Gross Domestic Product (GDP) growth. While emphasizing on the role of Investment (Grow Fixed Capital Formation), Human Capital (Employment and Labour Productivity Growth), and Trade Openness (Export and Import) by using time series data that covered from 1981 to The data analysis was preformed through econometric testing with Augmented Dickey-Fuller Test to check the stability of time series data. Johansen co-integration Test is employed to check whether Gross Domestic Product has empirically meaningful relationships with other variables or not? Our empirical findings reject the null hypothesis of no Co-integration and accept the co-integration relationship in our model. The Vector Error Correction Model and Granger causality test identify long-run equilibrium and short-run causality in GDP growth. The result of this research shows that GDP growth has long-run relationship with independent variables and short-run causality from Export, Import, and Employment but Grow Fixed Capital Formation and Labour Productivity Growth have no impact on GDP growth in short run. Keywords: Economic Growth, Human Capital, GDP, Unit Root Test, VECM and Causality 1. INTRODUCTION Economic growth is a measurable change that expands the output of the country in a given period. It is a process that increases the welfare of the nation (Osipian, 2009). The World Bank stated that Gross Domestic Product (GDP) is one of the macroeconomic indicators that measures the economic growth as an increase of national wealth that conventionally quantifiable in the percentage increase in gross domestic product (GDP) or gross national product (GNP). Economic development is a process of change that brings economic and social transformation (Thirlwall, 2006). Furthermore, economic development requires many processes that integrate physical capital, human capital and technological innovations (Walter, 1972). This paper reviews the source of economic growth that contributed to a current high economic growth of Ethiopia. Since 1991, the political and economic policies of Ethiopia had started to change radically and introduced liberal economic policies to encourage private enterprise and to attract foreign direct investment. Accordingly, Ethiopia registered impressive double digit economic growth for a decade and become one of the fastest growing economies in Africa (IMF, 2014; MoFED, 2014). This research is an attempt to answer what are the determinant factors of economic growth? It also evaluates other associated factors of economic growth. We use empirical data analysis and extended theories, and a quantitative approach on the data of economic growth. The first section of this paper reviews an overview of the GDP growth of Ethiopia. The second section focuses on the empirical analysis of time series data through econometrics models to evaluate the determinate of GDP growth. The data analysis uses Augmented Dickey-Fuller Test (ADF) to check the existence of unit roots and Johansen Co-integration Test to get the Co-integration between variables. The third part conducts Granger Causality Test to evaluate the determinate of GDP growth. The last part summarizes the empirical findings of the study and conclusion. 2. AN OVER VIEW OF ETHIOPIA S GDP GROWTH The economic growth of Ethiopia has been registering sustainable and vigorous growth for decades. The growth moment was driven by strong domestic demand, investment on infrastructural and economic liberalization. Moreover, the stable macroeconomic policies contributed for structural transformation from agricultural sectors to service sectors, the share of GDP shifted from low Australian Academy of Business Leadership 121

2 productive agriculture sectors to value added service sectors (IMF, 2011; McMilland Harttgen, 2014). Moreover, the government expenditure on infrastructure and human capital have been increasing dramatically, and also introduced the economic policies to encourage private sectors expansion. These have attracted world attention as investment destinations. Accordingly, Ethiopia registered impressive economic growth and become one of the fastest growing economies in the world. Ethiopia introduced market oriented economic policies to encourage private investment and also attracted foreign investment. This liberal economic policies promoted trade openness and provided tax incentive for export sectors and foreign direct investment. In addition, import substitution police introduced to transfer agricultural based economy to industrial based economy. These policies had contributed for macroeconomic structural transformations (Gedaand Berhanu, 2000; Rashid, Assefa, and Ayele, 2009; Berhanu and White, 2000). The classical economic theories supported that international trade has significant role in economic growth and create competitiveness through specialization (Jung and Marshall, 1985; Siddiqui, Zehra, Majeed, and Butt, 2008). Thus, Ethiopia implemented export led growth strategy to increase competitiveness and boost export that catalyzed GDP growth. The share of trade has increased sharply from 20% of GDP in 1990 to 45% of GDP in 2012 (MoFED 2014; WDI 2014). Investment is the main factor that derives economic growth (Janjili, 2011). Harrod-Domar growth model explained the rate of economic growth proportional to the rate of investment (Zhang, 2005). Saving accelerates investment that contributes to economic growth (Jappelli and Pagano, 1994). Furthermore, Solow s growth model emphasized the importance of physical capital for economic growth (Janjili, 2011). After introducing liberal economic policies, domestic saving has been growing from 9.7 percent of GDP in 1991 to 22.4 percent of GDP in At the same, gross fixed capital formation also has been growing proportional to saving from 14.5 percent of GDP in 1991 to 40.3 percent of GDP in (MoFED, 2014; World Bank, 2014; IMF, 2014). Several theoretical and empirical evidences show economic growth and human capital have positive relationship (Hafner, and Mayer-Foulkes, 2013; Hanushek, 2013). To understand the causal relationship between human capital and economic growth, it needs to develop a two-way approach of study (Hafner, and Mayer-Foulkes, 2013; Musai, and MEHRARA, 2013; Suri, Boozer, Ranis, and Stewart, 2011); therefore, the study employed Granger causality test. Economic growth recognized as the accumulation of human and physical capital, and increased productivity, arising from technological innovation (Lucas, 1998). The endogenous growth theory had given complementary theoretical support, which described human capital as the engine of growth through innovation (Barro, 2002). Human capital is a driving force of economic growth, which is the engine of growth (Benhabib and Spiegel, 1994). According Meraj (2013, pg.41) Adjustment in labour and capital is required to maintain long-run growth with the help of technological advancement in order to increase productivity. Solow growth theory also stresses the importance of physical capital that clearly defined the factors behind economic growth such as accumulation of capital, labour force and technology (Lucas, 1998). Economic growth was recognized as the accumulation of human and physical capital. Based on above argument, this paper evaluates the relationship between economic growth and the determinant factors such as Investment (Grow Fixed Capital Formation), trade openness (Export and Import), and Human Capital (Employment, and Labour productivity growth.) 3. METHODOLOGY The primary objective of the study is to empirically analyse the factors that affect economic growth by using time series data that covered from 1981 to The data analysis was employed with different econometric models. First, Augmented Dickey-Fuller (ADF) unit root test to measure the stability of the data. Followed by Johansen co-integration test to evaluate the series for integration of data and then Error Correction Model to identify the direction of causality in long-run equilibrium and short-run equilibrium of Vector Error correction model. Finally, Granger causality technique has employed to check the casual relationship among vectors. Before applying Co-integration test models, VAR Lag Order Selection Criteria is employed by using sequential modified LR test statistic (each test at 5% level), Final prediction error (FPE), Akaike Information Criterion (AIC), Schwarz Information Criterion (SIC), and Hannan-Quinn information criterion (HIQ). Lag Order Selection Criteria has employed to determine the optimal number of lag length. In addition, we preform Granger causality test by using Wald Statistic as well as we check GDP model whether it has any statistical problem by using the value of R-square and F-statistics (p-value). Then, we preform residual diagnostics test by using Breusch-Godfrey Serial Correlation LM test, Heteroscedasticity test of Breusch-Pagan-Godfrey, and Histogram Normality test of Jarque-Bera. This paper examines the dynamic relations between macroeconomic indicators of economic growth. The conceptual frame work of GDP and the model that proposed to evaluate the determinant of economic growth of Ethiopia are stated as GDP= +β 1 GFCF+β 2 EXPT+β 3 IMPT+β 4 EMPT+β 6 LPB+μ 122 Australian Academy of Business Leadership

3 Where, Gross Domestic Product (GDP), Gross Fixed Capital Formation (GFCF), Export (EXPT), Import (IMP), Employment (EMPT) and Labour Productivity Growth (LBP), = intercept and μ = error term, β= Coefficient. The A. Data Source The yearly time series data of Gross Domestic Product, Gross Fixed Capital Formation, Export, Import, Employment and Labour Productivity Growth collected from various sources, it covers from 1981 to The macroeconomic development data obtained from World Bank Development Indicators (WDI) Data base, The Ethiopian Ministry of Finance and Economic Development (MOFED), International Labour Organization (ILO), United Nations Educational, Scientific and Cultural Organization (UNESCO) and Central Statistical Agency of Ethiopia. B. Augmented Dickey-Fuller (ADF) Unit Root Test Before constructing an econometric model, Augmented Dickey-Fuller test is essential to preform to check unit root test before running Co-integration test in order to track autocorrelation problem. Because macroeconomic time series data have major problem for empirical econometrics that might cause spurious regressions, which create difficulty to measure regression results. The unit root tests are mainly a descriptive tool performed to classify the stability of time series data. Using ADF test exams whether the variable has unit root or not. Thus, ADF test determines order of integration of variables. The findings indicate all variables at level have unit root and became stationery at difference with trend and intercept. The null hypothesis assumed that a time series data appear to be non-stationary, which has unit root. The alternative hypothesis assumed that the date is stationer and reject null hypothesis (Dickey and Fuller, 1979). The following equation estimates the Augmented Dickey-Fuller model. x α+ βt+ πx + γ x + ε t= t 1 k t t i t i= 1 According to Fuller (1976), the null hypothesis is that x t =x (t-1) +ε t where ε t ~NID (0,σ 2 ). The notation NID (0,σ 2 ) symbolizes normally and independently distributed with mean zero and variance σ 2 or ε t the white noise error. The null hypothesis is H 0 :π=0[x t ~I(1)] against alternative hypothesis H a :π<0[x t ~I(0)]. The following Augmented Dickey-Fuller test selects an appropriate number of lag length by using automatic lag section criteria of Schwarz Information Criterion (SIC). The below Tables of ADF test result indicates all variables that examined are nonstationary at level, and accept the null hypothesis, which indicates all variables have unit root at 1%, 5% and 10% level. However, all variables became stationery at difference in Table 3 and Table 4 (trend and intercept) and the result reject null hypothesis at 1% and 5% level and accept the alternative hypothesis, thus, all variables become stationery at difference. Moreover, The above Table 1 shows Augmented Dickey-Fuller unit root test is for intercept only at 1% and 5% critical value, MacKinnon (1996) one-sided p-values. Table 1: Augmented Dickey-Fuller unit root test result for Intercept only at level Australian Academy of Business Leadership 123

4 The result of the abovetable 2 show Augmented Dickey-Fuller unit root test is for intercept only at 1% and 5% critical value, MacKinnon (1996) one-sided p-values. The above Table 3 indicates Augmented Dickey-Fuller unit root test results use intercept and trend at 1% and 5% critical value, MacKinnon (1996) one-sided p-values. Table 2: Augmented Dickey-Fuller unit root test result for Intercept only at difference Table 3: Augmented Dickey-Fuller unit root test results for Intercept and Trend at level Table 4: Augmented Dickey-Fuller unit root test results for Intercept and Trend at difference 124 Australian Academy of Business Leadership

5 The Table 4 indicates Augmented Dickey-Fuller unit root test results use intercept and trend at 1% and 5% critical value, MacKinnon (1996) one-sided p-values. C. Lag Length Selection Criteria Before performing Johansen Co-integration test and Vector correction error mode test, it requires to identify the number of optimal lag length by using VAR Lag Order Selection Criteria of sequential modified LR test statistic (each test at 5% level), Final Prediction Error (FPE), Akaike Information Criterion (AIC), Schwarz Information Criterion (SIC), and Hannan- Quinn Information Criterion (HIQ). It is necessary to determine the right lag length; because endogenous variables are highly sensitive to number of lag length. Thus, the lag selection criteria can select automatically an appropriate number of lag length. The finding indicates that all testing criteria are in favor of using two lag except SIC recommended to use one lag. Therefore, we use lag two as optimal lag length for Johansen Co-integration test and vector error correction model test. D. Johansen Co-Integration Test This paper employs testing for co-integration to evaluate long-run (equilibrium) relationships and short-run adjustment between variables. Johansen`s method checks if the GDP modelling has empirically meaningful relationships between vector. There are several tests of Co-integration. Among these methods, Engle and Granger (1987) formulated one of the first test of co-integration (or common stochastic trends), since then, the Engle-Granger (EG) has become a widely applied method of co-integration. In addition to EG long-run relationship approach, Johansen (1988, 1991) and Johansen and Juselius (1990) introduced a systems Table 5: Johansen Co-integration Test (Trace Statistic) Trace test indicates that there are four cointegratingeqn(s) at the 0.05 level. * denotes rejection of the hypothesis at the 0.05 level. **MacKinnon-Haug-Michelis (1999) p-values Table 6: Johansen Co-integration Test (Maximum Eigenvalue Statistic) Max-eigenvalue test indicates 4cointegratingeqn(s) at the 0.05 level. * denotes rejection of the hypothesis at the 0.05 level. **MacKinnon- Haug-Michelis (1999) p-values Australian Academy of Business Leadership 125

6 based approach to evaluate the existence of Co-integration among variables. Despite Johansen Co-integration test has weakness of the test on small sample size and sensitive to specification errors, it has theoretical advantage and methodological superiority (Sjö, 2008; Utkulu, 1997). Therefore, this paper employed Johansen Maximum Likelihood (ML) method to determine whether a stable long-run relationship (equilibrium) exists between the variables or not? The testing approach assumes that the system is integrated of order one. If there are signs of I (2) variables, we will transform them to I (1) before setting up the VAR. By using the difference operator Δ = 1 L, or L = 1 Δ, the VAR in levels can be transformed to a vector error correction model (VECM) (Sjö, 2008, p.14). Johansen Maximum likelihood approach identifies the number of cointegrating relationships between GDP and other variables. The Maximum Likelihood (ML) testing model construct based on trace test and Maximum Eigenvalue test. Where the null hypothesis is that the number of cointegrating vectors is r, against an alternative of (r+1) vector. The empirical model for this test is based on the following Trace statistics and Maximum Eigenvalue equations; as follow λ Max (, rr+ ) = T ln( 1 λr+ 1) λ Trace g () r = T ln( 1 λi) i=+ r 1 In additions, Max-eigenvalue test result in Table 6 also confirmed the same result as Trace statistic, which concluded that there are four Co-integration vectors in the model. Therefore, the null hypothesis of no Co-integration has been rejected and the study accepted the alternative hypothesis of existence of Co-integration among time series data. The finding determined that there is long run equilibrium (relationship) between the variables. E. Vector Error Correction Model (VECM) After evaluating the stability of vector for stationarity and unit roots by performing ADF test. The vector should be in levels and first differences. VECM will be employed once the variables integrated in the same order and cointegrated. Then we proceed to check whether a long-run equilibrium exists between variables. Furthermore, Wald statistics preformed to identify the direction of short-run Granger causality. The empirical model of VECM is represented by the following equation GDPt = + λzt n + β1 GDPt n + β2 GFCFt n + β3 EXPT + β4 IMPT + β5 EMPT + β 5 LBP + µ t Where, is the coefficient of error correction, and is error correction term (ECT), which is the lagged residual series of the cointegrating vector. denotes first differences and n is the optimal lag length determined by AIC and SC criteria and is the white noise term. The coefficient of cointegrated equation indicates the speed of adjustment toward long-run equilibrium, the coefficient must be negative and significant. denotes the coefficient of short-run equilibrium that measures Granger causality for Error correction model (ECM) of dependent variable. Coefficient parameters of error correction term are the speed of adjustment for the short-run imbalances. All the variables of vector error correction model are endogenously determined within GDP model and the empirical result indicates the coefficient is negative and significant with p-value of 3.9 percent. Therefore there are long-run causality from independent variable to GDP. Thus GFCF, EXPT, IMPT, EMPT and LBP have positive impact in GDP growth in long-run. Likewise, the Wald statistic result shows trade openness (Export and Import) cause GDP growth in short-run, which is in favour of liberalization theory. And also Employment has positive impact on short-run growth. Nevertheless, Gross Fixed Capital Formation, and Labour Productivity Growth have no short-run causality with GDP, which is against endogenous growth theory. The Granger causality test shows that GDP has bi-directional causality with export and import. However, the Granger causality findings surprisingly indicate that gross fixed capital formation, employment and labour production cause GDP, but the reverse is statically insignificant. In addition, export has bi-directional causality with import and gross fixed capital formation. Trade openness (import and export) cause labour productivity, due to the impact of knowledge spill over and positive externalities effect (learning by doing), but labour productivity does not granger cause import and export. Furthermore, this paper preformed residual diagnostics test by using Breusch-Godfrey Serial Correlation LM test, Heteroscedasticity test of Breusch-Pagan-Godfrey, and Histogram Normality test of Jarque-Bera. The study assess thoroughly the validity of time series regression data modelling assumptions. The finding indicates that there is no serial correlation and also there is no Heteroscedasticity. We therefore decide to accept the null hypothesis, which is desirable and expected. Hence, 126 Australian Academy of Business Leadership

7 the residual of GDP model has no autocorrelations and the regression model is homoskedastic. However, the Jarque-Bera test of normal distribution reject the null hypothesis. Therefore there is no normal distribution in the model. Finally, we evaluate whether the GDP model has statistical problem or not by checking the value of R-square and F-statistics (p-value). The finding concluded that there is strong R-square value ( ) and statically significant p-value ( ). 4. CONCLUSION This study has measured the determinate factors of GDP growth of Ethiopia by using Co-integration and Vector Error Correction model. The primary objective of the study is to determine the relationship between GDP growth and gross fixed capital formation, export, import, employment and labour productivity growth. The empirical findings show that the time series data has unit root at level and become stationery at difference. Moreover, the Co-integration test indicates that the series data is cointegrated in long run. Vector error correction model approach found evidence on the causality relationship between GDP and independent variable in long run. Likewise, the empirical result reveals that trade openness (export and import), human capital (employment and labour productivity growth) and physical investment (gross fixed capital formation) will cause GDP growth in long-run in Ethiopia. The Wald test causality findings surprisingly indicate that gross fixed capital formation does not cause GDP growth in short run, which is theoretically unexpected. In addition, Trade openness (import and export) cause labour productivity, this due to the impact of knowledge spill over and positive externalities effect (learning by doing), but the reverse is statically insignificant. REFERENCES Barro, R. J. (2002). Education as a determinant of economic growth (pp. 9-24). Benhabib, J., & Spiegel, M. M. (1994). The role of human capital in economic development evidence from aggregate cross-country data. Journal of Monetary economics, 34(2), Berhanu, B., & White, M. (2000). War, Famine, and Female Migration in Ethiopia, *. Economic Development and Cultural Change, 49(1), Dickey, D.A. and W.A. Fuller (1979) Distribution of the Estimators for Autoregressive Time Series with a Unit Root, Journal of the American Statistical Association, 74, p Fuller, W.A. (1976) Introduction to Statistical Time Series, John Wiley, New York. Geda, A., & Berhanu, N. (1960). The political economy of growth in Ethiopia. The Political Economy of Economic Growth in Africa, Hafner, K. A., & Mayer-Foulkes, D. (2013). Fertility, economic growth, and human development causal determinants of the developed lifestyle. Journal of Macroeconomics, 38, Hanushek, E. A. (2013). Economic growth in developing countries: The role of human capital. Economics of Education Review, 37, IMF Country Report No. 11/304. October Retrieved from (viewed on 20 th October 2015) IMF, Regional Economic Outlook Sub-Saharan Africa Fostering Durable and Inclusive Growth. Retrieved from (viewed on 29 th Jun 2015) Jangili, R. (2011). Causal relationship between saving, investment and economic growth for India what does the relation imply?. Jappelli, T., & Pagano, M. (1994). Saving, growth, and liquidity constraints. The Quarterly Journal of Economics, Jung, W. S., & Marshall, P. J. (1985). Exports, growth and causality in developing countries. Journal of development economics, 18(1), Johansen, S., & Juselius, K. (1990). Maximum likelihood estimation and inference on Co-integration with applications to the demand for money. Oxford Bulletin of Economics and statistics, 52(2), Johansen, S., & Juselius, K. (1990). Some structural hypotheses in a multivariate Co-integration analysis of the purchasing power parity and the uncovered interest parity for UK (No ). Lucas, R. E. (1998). On the mechanics of economic development. ECONOMETRIC SOCIETY MONOGRAPHS, 29, McMillan, M. S., & Harttgen, K. (2014). What is driving the African Growth Miracle? (No. w20077). National Bureau of Economic Research. Meraj, M. (2013). Impact of globalization and trade openness on economic growth in Bangladesh. Ritsumeikan Journal of Asia Pacific Studies (RJAPS), 32, MoFED, February Growth and Transformation Plan (GTP). Annual Progress Report for F.Y. 2012/13. Retrieved from com/documents/gtp-apr-2005-annual-progress-report pdf (viewed on 10 th October 2015) Musai, M., & Mehrara, M. (2013). The relationship between Economic Growth and Human Capital in Developing Countries. International Letters of Social and Humanistic Sciences, (05), Osipian, A. L. (2009). The Impact of Human Capital on Economic Growth: A Case Study in Post-Soviet Ukraine, Palgrave Macmillan. Rashid, S., Assefa, M., &Ayele, G. (2009). Distortions to Agricultural Incentives in Ethiopia. World Bank. Siddiqui, S., Zehra, S., Majeed, S., & Butt, M. S. (2008). Export-Led Growth Hypothesis in Pakistan Australian Academy of Business Leadership 127

8 Sjö, B. (2008). Testing for unit roots and Co-integration. Nationalekonomiskainstitutionen, Lindköpings Universitet. Suri, T., Boozer, M. A., Ranis, G., & Stewart, F. (2011). Paths to success: The relationship between HD and EG. World Development, 39(4), Thirlwall, A. P. (2006). Growth and development (Vol. 2). Palgrave Macmillan. Utkulu, U. (1997). How to estimate long-run relationships in economics: an overview of recent developments. Walter, R. (1972). How Europe Underdeveloped Africa. London. Bogle L ouverture. World Bank Development indicators (WDI)., (2014). Retrieved from (viewed on 30 th September 2015) Zhang, W. B. (2005). Economic Growth Theory: Capital, Knowledge, and Economic Structure. 128 Australian Academy of Business Leadership